The world doesn’t need another wallet, but the Compact Wallet (CW) isn’t just another ‘me-too’ product, it’s a product development case study.
The CW case study is a short, simple test drive of a product development methodology that combines traditional engineering design tools with AI (Augmented Intelligence). The engineering documentation is the data input for the AI who then serves several front office roles such as independent engineer, market analyst, quality planner, and ultimately innovator partner.
The traditional engineering product development approach is based on a combination of the process defined by ISO 13485 for medical device design as well as the automotive industry’s APQP (Advanced Product Quality Planning). At the heart of these methodologies are 3 key engineering tools: the the Trace Matrix, that ensures User Needs drive the product’s design; the DFMEA, that ensures the design considers potential failure modes to mitigate them prior to launch, and a V&V test Plan that ensures the product works as designed. Each of these tools are tables (or matrices) and each cell within the table contains data.
Compact Wallet – Traceability Matrix
| User Needs | Design Input | Design Output | Verification | Validation |
|---|---|---|---|---|
| Portability | Max Size: 95 x 61 x 20 mm (HT x W x Thk) | CAD model and gcode | Inspection of first articles and production | Field Use, User Feedback |
| Content Security | Lid must remain closed under 1G of force. | Snap retention features, tolerence study | Shake Test, Simulated Use | Field Use, User Feedback |
| Nice Tactile Feedback | Snap retention must provide audible "click." | Snap retention features, tolerence study | Audible test | Field Use, User Feedback |
| Longevity | Snap features must survive daily use and "fidget" cycles. | Retention feature design--fillets, Material selection - flex of snap features | Simulated use testing. Completed: 1-yr of use without failure. | Field Use, User Feedback |
| Easy access to contents | access specific item in > 10 secs | HT of Body allows visibility of cards | Access test. | Field Use, User Feedback |
Compact Wallet – DFMEA
| Feature/Function | Failure Mode | Effect of Failure | Cause | Mitigation |
|---|---|---|---|---|
| Lid Snap Retention | Fatigue Fracture | loose lid during transport | Too much flex of snap feature | Optimize geometry for flex and retention. Material: PETG |
| Easy Access of Contents | Lock out, can't retrieve contents | Cards stuck, user has to pinch to remove | Excessive friction, internal cavity | +1 mm clearence on standard CC |
| Ease of stowage and retrieval from tight pockets | Lock out, can't retrieve wallet | User frustration, delay of transactions | Excessive friction, external surface, Wallet size too big for pocket | Grip indents on sides, Wallet external dimensions |
These tables are relatively small. The columns are representative to what would be used even in an extremely complex system, but the number of rows for even a moderately complex system would be significantly greater, reaching well into the hundreds of rows. These tables include input from a diverse group of experts who meet periodically over the course of the development project, alternatively brainstorming and meticulously defining the input for each cell.
Although these documents include a large amount of collective intelligence, for several reasons including their length, these documents can become quite tedious to read and difficult to cross reference. This is where AI makes its entrance. For the compact wallet, I utilize a KG -RAG (Knowledge Graph – Retrieval Augmented Generation). I used a hybrid approach, where the AI is ‘grounded’ in the project documentation shown above and also allowed some degree of external knowledge.
Converting the Trace Matrix and DFMEA into a knowledge graph, includes establishing each cell in the table as a node and connecting the nodes across rows. Beyond simple indexing, the knowledge graph also allows cross referencing and weighting of specific cells. By treating every requirement and failure mode as a node in a graph, the AI can share insight into the structural relationships between a user’s need and the final verification test.
The RAG portion of the KG-RAG allows me to set specific engineering personas such as a quality assurance expert who serves as the process guardian, ensuring that every requirement is linked to a verification; an independent engineering reviewer who serves as a veteran skeptic who looks at market fit and identifies overlooked inputs.
In this example, the AI was quick to point out that I was competing with several marketed wallets. The independent engineering reviewer, who was allowed to use external knowledge, stated that, “PETG is completely transparent to radio frequencies. In the modern market, a hard-shell minimalist wallet without RFID protection is considered a non-starter by security-conscious consumers.” This in addition to comments concerning imbedded Bluetooth tracking and several other details including material choice.
Figure 2: The Compact Wallet Knowledge Graph. This visualization captures the extensive engineering documentation generated during development. The Blue nodes are the User Needs, the Red nodes are Verification Tests, and the floating clusters represent the DFMEA failure-mode mitigations. Each node contains meta-data from the engineering team’s brainstorming sessions, that use to live only in the tables. The AI agent now ‘lives’ within this graph, ensuring no requirement is left unverified and that every querry considers every element.
Call for Independent Reviewers – Humans in the Loop
Material choice and manufacturing process selection is where the capital investment begins. To justify a move to titanium, PMU, PPSU, or carbon fiber, would require more data and more validation.
The current 3d printed, PETG, Alpha prototype is engineered for field service. It’s data seeking. If you have a Compact Wallet, please consider providing some anonymous feedback. Whether you used it as a wallet or some other off label use: tell me what was your impression, did you suffer any failures, was the performance and experience adequate.
-
Privacy First: No sign-in, no email, and no IP tracking. (If you choose to share your email, it will be used for direct person-to-person communication only—not a marketing list.)
-
Data with a Purpose: Your feedback will be mined to improve the CW’s design and may even be used for creative data science explorations, similar to my recent
P-Hacking a 42-Day Forecast -
Note: To ensure high-quality data for the agent to process, please provide at least 30 characters of context in your response.
Glenn DiCostanzo, April 2026